# Three analytical questions this dataset can answer The dataset holds daily price and volume history for 152 semiconductor stocks, from 2012-01-02 to 2025-08-04 (422,729 rows). 99,984 of those rows (23.7%) fall after ChatGPT's launch, which gives a natural before/after split. All figures below come from the profiling queries I ran (steps 0–12). ## 1. Did the AI surge change how semiconductor stocks behave, and for which companies? - **Comparison:** Compare each company's price trajectory, daily volatility and trading volume before and after the ChatGPT launch. - **Why it's interesting:** The "AI winners" story is usually told about a few names. With 152 companies you can test whether the shift is broad across the sector or concentrated in a handful of stocks. - **Baseline to beat:** Across the whole history, the average daily high-low range is about 3.69% of the close, the median close is about $35.57, and the median daily volume is 898,500 shares. Per-company before/after deltas can be measured against these. - **Decision it serves:** Which stocks to overweight or underweight, and whether the surge justifies entering or exiting a position. ## 2. Which stocks carry the most risk per unit of return, and has that changed? - **Comparison:** Rank companies by volatility (the high-low range as a share of close) against their price growth since 2012, then compare the rankings before and after 2022. - **Why it's interesting:** The 3.69% average range hides large differences between stocks. Some may have had big gains at a modest risk cost, while others have only added volatility. - **Decision it serves:** Position sizing and choosing which names belong in a sector basket or benchmark. ## 3. When does unusual trading volume signal something, and what follows it? - **Comparison:** Flag days when a stock's volume is far above its own norm, then look at the price moves in the days after. - **Why it's interesting:** Volume is highly skewed. The mean is about 8.47M shares, nearly ten times the 898,500 median, and the maximum is about 3.69 billion in a single day. Volume spikes are clearly present, but they need a per-stock baseline to be meaningful. - **Caveat:** 4,888 rows have zero volume. They should be excluded, or checked to see whether they are halts or gaps in the data, before any spike analysis. - **Decision it serves:** Timing entries and exits. ## Caveats - **Observational only:** This is market data with no experiment, so the AI effect can be described but not isolated from broader market moves. Benchmarking against the sector as a whole helps. - **Company count:** The dataset card says about 149 companies, but the query counts return 152. - **Not yet run:** I profiled the dataset's overall shape, but I have not yet run these three analyses. I can run any of them next, starting with the per-company before/after comparison.